Yukon Egg Analysis
Distribution Visualizations
Egg mass distributions
First, I wanted to visualize the average egg mass distribution by collection site.
We can see a general increase in average egg size as the fish move up the river. Most sites have a generally normal distribution, Chena does not but there are only 3 fish. I decided to remove Chena and Salcha from downstream analyses. I can easily add them back in.
Canada-bound fish (2023)
Here I have filtered to only include fish genetically identified as Canada-origin from the PIST group, I have also included fish collected at RARA, FOYU, and WH.
Take home messages from the analyses below, as fish move upriver:
- Egg mass increases
- Egg thiamine concentration decreases
- Egg % moisture decreases, then increases
- Egg total moisture content increases
- Egg % lipid (dry) decreases
- Egg total lipid increases from PIST
Average Egg Mass
Distribution
The average egg mass distribution is largely unchanged when non-Canada-bound fish are filtered out.
Violin Plot
Average egg mass is significantly different at each site along the river. Eggs are getting larger as the fish proceed towards Whitehorse.
These data meet the assumptions of an ANOVA test. The data are mostly normally distributed, there is homoscedasticity of variance (Levene’s test p > 0.05), and the observations are independent of each other. The observations are independent because we are only looking at Canada-bound fish, if we were looking at fish assigned to different “stocks” we would need to account for the random effect of “stock”.
We aren’t 100% sure all the fish from RARA and FOYU are Canada “stock” we need to be sure of this.
Egg Thiamine Concentration
Distribution
Distributions become more normal as fish migrate upriver and egg thiamine concentration drops.
Violin Plot
Egg thiamine concentration decreases as fish migrate up the river. Fort Yukon is not significantly different from Whitehorse.
Data are not normally distributed in each group and variance is not homoscedastic, so ANOVA could not be used. Instead I tested the difference between groups using the non-parametric Kruskal-Wallis test and used a post-hoc Dunn’s test with Bonferroni adjustment. The Kruskal-Wallis test compares the group medians (mean ranks) rather than the means like an ANOVA. Because the variance of each group is not homoscedastic, all we can confidently say is that the distributions of the different groupings are different.
Egg Percent Moisture
Distribution
Violin Plot
Here, we are comparing group medians because the variance is homoscedastic, but the data are not normally distributed. Again we used a Kruskal-Wallis test and a post-hoc Dunn’s test.
Egg Total Moisture
Distribution
Violin Plot
Again the data are not normally distributed, but we can see an increasing trend in total moisture content as the fish move up the river. The variance is homoscedastic so we are comparing the medians of the different groups.
Egg % Lipid (dry)
Distribution
Violin Plot
Egg percent lipid (dry) decreases as fish move upriver. Not normally distributed, variance is homoscedastic (Kruskal-Wallis tests differences of medians).
Egg Total Lipid Content
Distribution
Data pass tests for normality and homoscedasticity, testing differences of means with ANOVA.
Violin Plot
Lipid content increases from PIST, but all other sites are do not have significantly different mean lipid content values.